A Statistical Method for Method Comparison Studies When Outcomes Can Only Assume Monotone

• Activity Number: 596
• Type: Topic Contributed
• Date/Time: Wednesday, August 12, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Medical Devices and Diagnostics
• Author(s): Yuqing Tang* and Meijuan Li and Jincao Wu
• Companies: FDA and FDA and FDA
• Keywords: method comparison ; monotone ; real example

Abstract

When a new diagnostic device with continuous output is proposed, we can evaluate its measurement accuracy by comparison with an established device (or comparator) with the same indications for use as the new device by so-called method comparison study. It involves the comparison via paired results from the new device and comparator. Statistical methods as well as CLSI guidance are available for method comparisons when two devices measure the same quantity and have acceptable linearity over their measuring ranges. However, problem arises when the new device and comparator do not measure the same quantity or they measure the same quantity, but their outcomes do not have acceptable linear relationship with true unknown quantity. In either of these two cases, the new device is not expected to agree well with the comparator and consequently, the existing methods for assessing agreement are not applicable. We propose to develop a novel statistical approach to address this problem. A real example will be used to illustrate the application of our approach.